Breathing Pattern Monitoring by Using Remote Sensors

Area of Science:

• Physiological monitoring
• Biomedical engineering
• Machine learning for healthcare

Background:

• Respiration is closely linked to physiological processes, making breathing pattern analysis crucial for health assessment.
• Altered breathing patterns can indicate underlying pathophysiological changes, enabling early detection of health issues.

Purpose of the Study:

• To develop and evaluate a contactless system for continuous and unobtrusive monitoring and classification of breathing patterns.
• To create robust respiratory signal extraction algorithms for differentiating various breathing patterns.

Main Methods:

• Volunteer subjects followed a predefined breathing protocol.
• Color and thermal cameras recorded breathing patterns.
• Developed and compared respiratory signal extraction algorithms.
• Trained a multiclass support vector machine (SVM) using extracted respiratory features.

Main Results:

• A novel algorithm robustly extracted multiple respiratory features.
• The system achieved 95.79% accuracy in distinguishing between different breathing patterns.
• The developed algorithm effectively differentiated a wide range of respiratory patterns, including complex ones.

Conclusions:

• Contactless monitoring of breathing patterns is feasible and accurate for automated health assessments.
• The developed system and published dataset can advance research in non-invasive respiratory monitoring and classification.
• Machine learning, particularly SVMs, shows significant potential in analyzing complex breathing patterns.